Arno Onken
@arnoonken.bsky.social
Lecturer at the University of Edinburgh, interested in probabilistic and machine learning methods for modeling and analyzing neural activity.
Reposted by Arno Onken
The model also works with datasets containing a few hundred neurons from different animals and laboratories. There is more good stuff in the appendix of the paper and the code repository!
Paper: www.biorxiv.org/content/10.1...
Code and model weights: github.com/bryanlimy/Vi...
7/7
Paper: www.biorxiv.org/content/10.1...
Code and model weights: github.com/bryanlimy/Vi...
7/7
Movie-trained transformer reveals novel response properties to dynamic stimuli in mouse visual cortex
Understanding how the brain encodes complex, dynamic visual stimuli remains a fundamental challenge in neuroscience. Here, we introduce ViV1T, a transformer-based model trained on natural movies to pr...
www.biorxiv.org
September 19, 2025 at 12:37 PM
The model also works with datasets containing a few hundred neurons from different animals and laboratories. There is more good stuff in the appendix of the paper and the code repository!
Paper: www.biorxiv.org/content/10.1...
Code and model weights: github.com/bryanlimy/Vi...
7/7
Paper: www.biorxiv.org/content/10.1...
Code and model weights: github.com/bryanlimy/Vi...
7/7
Reposted by Arno Onken
We sincerely thank Turishcheva & Fahey et al. (2023) for organising the Sensorium challenge(s!) and for making their high-quality, large-scale mouse V1 recordings publicly available, which made this work possible!
6/7
6/7
September 19, 2025 at 12:37 PM
We sincerely thank Turishcheva & Fahey et al. (2023) for organising the Sensorium challenge(s!) and for making their high-quality, large-scale mouse V1 recordings publicly available, which made this work possible!
6/7
6/7
Reposted by Arno Onken
We compared our model against SOTA models from the Sensorium 2023 challenge and showed that ViV1T is the most performant while being more computationally efficient. We also evaluated the data efficiency of the model by varying the number of training samples and neurons.
5/7
5/7
September 19, 2025 at 12:37 PM
We compared our model against SOTA models from the Sensorium 2023 challenge and showed that ViV1T is the most performant while being more computationally efficient. We also evaluated the data efficiency of the model by varying the number of training samples and neurons.
5/7
5/7
Reposted by Arno Onken
Moving beyond gratings, we used ViV1T to generate centre-surround most exciting videos (MEVs) via the Inception Loop (Walker et al. 2019). Our in vivo experiments confirmed that MEVs elicit stronger contextual modulation than gratings, natural images and videos, and most exciting images (MEIs).
4/7
4/7
September 19, 2025 at 12:37 PM
Moving beyond gratings, we used ViV1T to generate centre-surround most exciting videos (MEVs) via the Inception Loop (Walker et al. 2019). Our in vivo experiments confirmed that MEVs elicit stronger contextual modulation than gratings, natural images and videos, and most exciting images (MEIs).
4/7
4/7
Reposted by Arno Onken
ViV1T also revealed novel functional features. We found new properties of contextual responses to surround stimuli in V1 neurons, both movement- and contrast-dependent. We validated this in vivo!
3/7
3/7
September 19, 2025 at 12:37 PM
ViV1T also revealed novel functional features. We found new properties of contextual responses to surround stimuli in V1 neurons, both movement- and contrast-dependent. We validated this in vivo!
3/7
3/7
Reposted by Arno Onken
ViV1T, only trained on natural movies, captured well-known direction tuning and contextual modulation of V1. Despite no built-in mechanism for modelling neuron connectivities, the model predicted feedback-dependent contextual modulation (including feedback onset delay!) (Keller et al. 2020).
2/7
2/7
September 19, 2025 at 12:37 PM
ViV1T, only trained on natural movies, captured well-known direction tuning and contextual modulation of V1. Despite no built-in mechanism for modelling neuron connectivities, the model predicted feedback-dependent contextual modulation (including feedback onset delay!) (Keller et al. 2020).
2/7
2/7
Thank you for the correction! It seems the pictures are showing the Anatomy Lecture Theatre whereas Burke was dissected in the Usha Kasera Lecture Theatre, which was known as the Old Anatomy Lecture Theatre. However, Burke's skeleton is on exhibition right next to the theatre shown in the pictures.
January 29, 2025 at 12:46 PM
Thank you for the correction! It seems the pictures are showing the Anatomy Lecture Theatre whereas Burke was dissected in the Usha Kasera Lecture Theatre, which was known as the Old Anatomy Lecture Theatre. However, Burke's skeleton is on exhibition right next to the theatre shown in the pictures.
Yes, indeed! It is a lecture theatre of historical significance. In 1829, William Burke of murderers Burke and Hare was dissected in that lecture theatre. Burke and Hare murdered 16 people and sold their corpses for anatomy lectures.
January 29, 2025 at 10:25 AM
Yes, indeed! It is a lecture theatre of historical significance. In 1829, William Burke of murderers Burke and Hare was dissected in that lecture theatre. Burke and Hare murdered 16 people and sold their corpses for anatomy lectures.